Anonymous | Login
Project:
RSS
  
News | My View | View Issues | Roadmap | Summary

View Issue Details[ Jump to Notes ] [ Issue History ] [ Print ]
ID
0052457
TypeCategorySeverityReproducibilityDate SubmittedLast Update
defect[Openbravo ERP] A. Platformmajorrandom2023-05-16 10:292023-05-16 10:31
ReportergorkaionView Statuspublic 
Assigned ToTriage Platform Conn 
PriorityhighResolutionduplicateFixed in Version
StatusclosedFix in branchFixed in SCM revision
ProjectionnoneETAnoneTarget Version
OSAnyDatabaseAnyJava version
OS VersionDatabase versionAnt version
Product VersionSCM revision 
Merge Request Status
Review Assigned To
OBNetwork customerGold
Web browser
ModulesCore
Support ticket64183
Regression level
Regression date
Regression introduced in release
Regression introduced by commit
Triggers an Emergency PackNo
Summary

0052457: On cluster environments processing node can be switch even if current node is alive

DescriptionOn a cluster environment with 2 applications nodes.
Node A is handling Import Entries.
Node B does pings every second to check if Node A is alive by checking a database timestamp.

If for some reason Node A is slow updating the database timestamp the node B can take ownership of the Import Entries even if node A is still alive.

The ping is done every 10s and only allows 1 extra second to consider node A down. So in case Node A fails to update the reference timestamp in 11 seconds the nodes are switches.

Since the node A is not really down this causes that some Import Entries are processed in both application nodes.

The handling process should allow more pings to fail before executing the switch of nodes to allow some punctual slowness in node A.
Steps To ReproduceForce node A to delay the reference timestamp update for 12 seconds.
Proposed SolutionIncrement the extra time allow 1 or 2 pings to fail before switching handling nodes.
TagsNo tags attached.
Attached Files

- Relationships [ Relation Graph ] [ Dependency Graph ]
duplicate of defect 0052445 closedAugustoMauch The margin when checking if we need to replace a cluster instance is too small 
related to defect 0052456 closedcaristu Import Entry post process hook not executed when processing node is changed 
related to defect 0052455 closedcaristu Import Entry processed twice when processing node is changed 

-  Notes
There are no notes attached to this issue.

- Issue History
Date Modified Username Field Change
2023-05-16 10:29 gorkaion New Issue
2023-05-16 10:29 gorkaion Assigned To => Triage Platform Conn
2023-05-16 10:29 gorkaion OBNetwork customer => Gold
2023-05-16 10:29 gorkaion Modules => Core
2023-05-16 10:29 gorkaion Support ticket => 64183
2023-05-16 10:29 gorkaion Triggers an Emergency Pack => No
2023-05-16 10:30 gorkaion Relationship added related to 0052456
2023-05-16 10:30 gorkaion Relationship added related to 0052455
2023-05-16 10:31 gorkaion Relationship added duplicate of 0052445
2023-05-16 10:31 gorkaion Status new => closed
2023-05-16 10:31 gorkaion Resolution open => duplicate

Copyright © 2000 - 2009 MantisBT Group
Powered by Mantis Bugtracker